Datalicious Depths of Delectable Data

Want to jump to a specific question?

Which results are the most common?

See, if Open Psychometrics recorded the respondents’ results, it would be quite quick to answer this question. Unfortunately, they didn’t, so we’re working with a bunch of numbers!

This is what the data looks like:

I’ve selected the first 6 columns of the data set… and it’s not too bad! It’s a little intimidating to think that there are over 1 million more rows, but we’ve gotta start somewhere >:)

We start by calculating the scores for the ‘Extroversion’ personality trait. We can do this by summing up the scores of the 10 EXT (standing for EXTroversion) questions. However, it should be noted that answering a ‘4’ or ‘5’ on a EXT question doesn’t always mean that a user is more extroverted. For example, answering 5 for “I don’t talk a lot” vs “I am the life of the party” would significantly affect your results.

So, we’ll need to make sure to add and subtract scores as necessary, depending on how the questions are keyed. I’ve arbitrarily made the decision to ‘positively key’ questions symbolising extroversion (meaning that I’d add points to their total score if they showed signs of extroversion), and ‘negatively key’ questions symbolising introversion (subtracting points from total score). See the table below to better understand this keying:

Table 1: EXT Question Key
Questions Key
I am the life of the party. \(+\)
I don’t talk a lot. \(-\)
I feel comfortable around people. \(+\)
I keep in the background. \(-\)
I start conversations. \(+\)
I have little to say. \(-\)
I talk to a lot of different people at parties. \(+\)
I don’t like to draw attention to myself. \(-\)
I don’t mind being the center of attention. \(+\)
I am quiet around strangers. \(-\)

Hence, after adding/subtracting their points from each question, any positive total score (> 0) will mean that the user receives an ‘S’ for Sociable, while a negative score (< 0) would result in an ‘R’ for reserved.1 If their score is 0, then they’d receive an ‘X’, as their results since they’re perfectly in-between S and R, and will remain inconclusive.

Below is a sample table of results, where extletter (their final letter result) and extscore (the sum of their points) are calculated from the raw data.


We can also visualize our results into a graph:

The same processes will be applied to the other personality traits, with their graphs located below:

Now, we can combine everybody’s letters from each personality traits to form their final result. Then we can determine which results are the most common within this data set.

In fact, we can see the 25 most common results below:


WOOO! Out of the 243 possible answers, congratulations to the __OAI family for dominating the top 4 spots!

If you’d prefer to see a more numerical representation, I’ve provided a table with the exact number of the top 10 results and the average score of each personality trait:2

Table 2: The 10 Most Common Results
Results Amount of Responses Percentage of People
SCOAI 159575 15.716379
RLOAI 121251 11.941888
SLOAI 100841 9.931727
RCOAI 92548 9.114958
RLUAI 70642 6.957459
SLUAI 56870 5.601068
SCUAI 41138 4.051640
RCUAI 26379 2.598041
RLOEI 23547 2.319120
RCOEI 22333 2.199554
Table 3: Average Score for Each Personliaty Trait
Extroversion Neuroticism Conscientiousness Agreeableness Openness
avgscore 0.72 0.16 3.813333 7.48 8.986667

Let’s try to analyze why these results happened.

Extrovertism

Extrovertism is the most balanced trait, as there seems to be a nice mix of S(ocial) and R(eserved) – representative of the larger population. I’d assume this is because questions in this trait were very straight-forward (e.g. “I start conversations”), and users likely display/are conscious of these behaviors in their everyday life. Hence, users will have a more ‘objective’ viewpoint, leading to more accurate results.

There are also about ~40k more introverts compared to extroverts. Maybe the S’s aren’t as likely to spend 20 minutes on an online test, compared to R’s?

Neuroticism

Once again, a solid distribution of results between C(alm) and L(imbic), mirroring what I’d expect the general population to be like. Similar to the extrovertism section, the questions tend to be straightforward and there’s relatively less self-bias within the trait, as I believe that many people are conscious of their experience with their emotions emotions, and whether or not they control them.

There are ~60000 more limbic people compared to calm. I’m not really surprised - people are more stressed3 and more sad4 than we’ve ever been. Furthermore, emotional regulation is a result of introspection and reflection. As our society slowly moves towards constant stimulation and excess, it’s no wonder we struggle to achieve emotional stability.

Conscientiousness, Agreeableness, and Openness

With these three personality traits, we start to see a much larger gap between the possible letters. There are significantly more people that are organized, agreeable, and inquisitive, relative to unstructured, egocentric, and non-inquisitive.

I definitely don’t think these results genuinely reflect the general population, despite the fact that I have no evidence stating so. If I had to speculate, I’d say that this data is just a quite biased, leading to skewed results.

  1. We’re just a little bit tainted

No matter how hard we try, we can never be completely objective. Many struggle to see things beyond ourselves and to judge every action we take from a neutral standpoint.

In normal, everyday life, this is not too significant of an issue. A little bit of subjectivity never hurt anybody! However, when we’re trying to objectively analyze and record data about ourselves, our lack of objectivity can lead to delusional answers - espicially since this is an self-administrated quiz.

  1. We’re fragile creatures

For these three categories, some traits are ‘better’ to have than others. In this society, it is almost ALWAYS better if you’re organized, agreeable, and inquisitive, rather than the opposite. In fact, it’s almost an insult if we AREN’T these things.5

For example, people tend to be more agreeable and people-pleasing, since it shows that you’re friendly and nice. Because you exhibit these behaviours people will tend to ‘like’ you more (relative to being confrontational and assertive for your own needs). Similar comparisons can be made for being conscientiousness, especially as our society moves towards maximizing productivity and ‘grind culture’. People NEED to be conscientious and self-disciplined, or else you won’t be successful. Or, in relation to inquisitivity, people don’t want to be seen as rigid and unwilling to try new things. Society will call them scaredy-cats, boring, or tell them that you’re ‘bringing the mood down’.

It doesn’t help that we don’t like being ‘bad people’ – we constantly attempt to justify our actions to be good and perceive ourselves in a positive light. We are rarely the villains in our own stories; it’s always the other person doing something wrong, or doing something worse than we did, or they started the ordeal.

However, when we can’t be objective, we can confuse what we actually do, with what we wish we did. This is going to lead us to choose answers corresponding to those ‘better’ traits of self-discipline, agreeableness and inquisitiveness, instead of objectively saying we have the ‘worse’ traits of unstructuredness, egocentrism and un-inquisitiveness.

  1. We’re just built for this (for Openness)

There’s also the classic case of sampling bias. The people who take this quiz are probably curious about psychology, want to know more about themselves, and are willing to try something new. These are typically the types of people who are inquisitive (matching with the I letter) instead of non-inquisitive (matching with the N letter).

Sheer Statistical Impossibility…?

When analyzing this data, it seems.. strange that so many __OAI types are represented. In fact, it feels weird that 15%(!!!!) of people were the EXACT SAME TYPE, despite ALL POSSIBLE COMBINATIONS!

So, let’s see how it compares to the theoretical data. Using the data6 from SimilarMinds, we can see how our data stacks up.

Table 4: Theoretical Data
Results Percentage of People
SCOAI 3.4
RLOAI 2.7
SLOAI 2.5
RCOAI 3.5
RLUAI N/A
SLUAI 3.4
SCUAI 4.1
RCUAI N/A
RLOEI N/A
RCOEI N/A
Table 4: Experimental Data
Results Percentage of People
SCOAI 15.716379
RLOAI 11.941888
SLOAI 9.931727
RCOAI 9.114958
RLUAI 6.957459
SLUAI 5.601068
SCUAI 4.051640
RCUAI 2.598041
RLOEI 2.319120
RCOEI 2.199554

Looking at these tables, there’s a clear discrepancy between theoretical and experimental data, where only SCUAI matches the theoretical results. It’s clear that certain results are far more represented in our data set than the average value of the general population.

I believe that most of it can be explained by the previous analysis: Wrong results might occur because of biases and societal norms (cementing the last two letters to be A and I, and increasing the likelihood of an O vs an U), in addition to the natural disposition of respondents.

Honestly, I’m not too sure why 15% of people are SCOAI, maybe people are being influenced to answer what they WANT to be like, rather than what they actually are. My best guess is that SCOAI seems like one of the most ‘socially-successful’ results, since extroverted, stable, organized, agreeable, and inquisitive are all attributes that allow people to thrive in this society. They tend to have larger social circles, better interpersonal relationships and private lives, and the ability to study and work hard to get to positions they want academically and corporately.

So, my guess is that many respondents are not actually SCOAIs, and they’re not answering these questions objectively. That’s my best guess!7

Also, just to provide a comprehensive review of the results, here’s a list of the most uncommon ones!

Table 5: Most Uncommon Results
Results Amount of Responses
XXUXN 3
SXXXN 5
XLXXN 5
XXUEX 5
SCXXX 6
XCXEX 6
XXOXN 7
SXOXX 9
XCUXX 9
XCXXN 9

No surprise, it’s a lot of results with X’s!

At first, I was surprised that ‘XXXXX’ didn’t appear, since I’d assume it (theoretically) is the most unlikely.8 However, I wouldn’t be surprised if only 5% of XXXXX respondents were genuinely XXXXX, while the other 95% of people who got the result just kept clicking 3 (Neutral) for every question, or had some kind of game to see if they could get the (theoretically) super rare XXXXX.

Do results vary between countries?

This data contains 224 unique ISO country codes.9 Let’s dig through this data - a fun bit of stalking!

We can see that the majority of data came from the US, with a whopping total of 546403 respondents. Trailing (very far) behind, we also have Great Britain (GB), Canada (CA), and Australia (AU).

This is likely because this quiz is in English, and will generally cater towards countries with English as their primary language. In addition, Google’s SEO (search engine optimization) algorithm is also affected by location, and will rank websites by their proximity to the user.10 So, it’s possible that Open Psychometrics is located in America,11 and when Americans search up “Big Five Personality Test”, this would be the first quiz that shows up.12

But, we’re not really concerned on WHERE people are taking the quiz. Instead, we only care about how it affects the responses. Hence, we’re going to start by finding the most common results for America, Great Britain, Australia, and the Philippines – these location have the most amount of results while being from different continents – and see how they compare to one another.

Table 6: America
Results Percentage
SCOAI 16.855325
RLOAI 12.283425
SLOAI 10.141050
RCOAI 9.943943
RLUAI 6.578112
SLUAI 5.344041
SCUAI 4.044268
RCUAI 2.546472
RCOEI 2.150610
RLOEI 2.142924
Table 6: Great Britian
Results Percentage
SCOAI 13.129918
RLOAI 11.623821
SLOAI 10.149258
RLUAI 8.353355
SLUAI 7.272208
RCOAI 6.776683
SCUAI 4.606883
RLOEI 2.660820
RCUAI 2.575230
RLUEI 2.386029
Table 6: Australia
Results Percentage
SCOAI 16.546072
RLOAI 11.279233
SLOAI 9.696182
RCOAI 9.020588
RLUAI 6.534080
SLUAI 5.954427
SCUAI 4.309414
RCUAI 2.372577
RLOEI 2.072756
RCOEI 2.046772
Table 6: Philippines
Results Percentage
RLOAI 15.105557
SCOAI 11.699501
SLOAI 10.540636
RCOAI 8.726760
RLUAI 5.869905
SLUAI 3.305285
RLOEI 2.216960
RLOAN 2.191767
XLOAI 1.889454
RCUAI 1.627450

From the above table, it’s pretty clear that countries have very similar trends, with almost identical top 7 results between all the countries. However, there are slight differences:

Generalizing the World

We can also plotting the trait averages on a world map, and then analyze the results.

A note on the data: Locations with less than 10 responses have been omitted from the map data, as they often significantly skew the scale of the maps. This removed 58 locations off the map, with Africa losing a pretty big chunk of their land.

(#fig:extscores world map)Diamond sadasds

Higher scores are correlated with extroversion, lower scores are correlated introversion.

Table 7: Most Extroverted
region ExtScores
Cuba 2.708333
Greenland 2.382353
Rwanda 2.093750
Ethiopia 1.342960
Afghanistan 1.018519
Norway 0.989608
Table 7: Least Extroverted
region ExtScores
St. Kitts & Nevis -6.333333
Sudan -5.466667
St. Lucia -5.190476
Åland Islands -4.733333
Guyana -4.543478
Bhutan -4.500000

There doesn’t seem to be an overarching trend in the world, but there are several generalizations.

It’s quite interesting that the sterotype is that Americans are more extroverted and loud, while Asians are more introverted and quiet. Doesn’t seem to apply to this dataset!

Higher scores are correlated with being calm, lower scores are correlated being limbic.

Table 8: Most Calm
region estScores
Suriname 4.697674
Eswatini 3.461539
Ethiopia 3.451264
Cape Verde 3.181818
Cuba 3.000000
Papua New Guinea 2.791667
Table 8: Least Calm
region estScores
Jersey -4.476191
Guernsey -4.044444
Syria -3.875000
Samoa -3.727273
Algeria -3.234310
Belize -3.125000

It seems like the south-eastern part of the world is less limbic, specifically Africa and East Asia. I find it funny that China specifically is seen to be relatively ‘calm’, as they are notorious for bad work-life balances,13 the ‘lie down’ movement14 – factors that would propagate negative moods and emotional instability.

Yet, it’s also reasonable to say that Chinese people are accustomed to high stress after dealing with academic pressures.15 Hence, they might have learned to better regulate their emotions.

Higher scores are correlated with organization and contentiousness, lower scores are correlated carelessness.

Table 9: Most Contentious
region csnScores
Ghana 6.922794
Papua New Guinea 6.708333
Grenada 6.291667
Rwanda 6.281250
Uganda 6.127517
Kenya 6.111744
Table 9: Least Contentious
region csnScores
Bhutan -1.0000000
Bolivia 0.0575342
Libya 0.1764706
Angola 0.3571429
Åland Islands 0.6000000
Paraguay 0.6975089

For conscientiousness, Africa is lit up like a Christmas tree! North America is also pretty light. On the other hand, South America is quite dark.

I find it interesting that Asia, which is known for their disciplined schedules and focus, seems to be quite average. Furthermore, Sub-Saharan Africa seems to be incredibly conscientious - yet, there’s a direct link between low conscientiousness and poverty.16. Is this just another instance of self-bias?

Higher scores are correlated with agreeableness, lower scores are correlated egocentrism.

Table 10: Most Agreeable
region agrScores
Papua New Guinea 11.04167
Rwanda 10.56250
Cameroon 10.45455
St. Lucia 10.42857
Cuba 10.20833
Tanzania 10.17442
Table 10: Least Agreeable
region agrScores
Madagascar 1.500000
Åland Islands 1.533333
Bhutan 3.428571
Poland 3.818162
Cape Verde 3.909091
Belarus 3.927711

I had to double check twice to make sure I didn’t accidentally duplicate the conscientiousness graph, as they look almost identical.

I am, once again, not surprised that everybody thinks they’re agreeable. When looking at the scores on the ‘least agreeable’ table, not a single country is willing to admit they’re… just a tiny bit egotistical! Alas, the bias goes deep.

(#fig:opn world map)Diamond Prices

Higher scores are correlated with inquisition, lower scores are correlated traditionalists.

Table 11: Most Inquisitive
region opnScores
Madagascar 12.50000
Cuba 11.37500
Seychelles 11.36364
St. Lucia 11.04762
Armenia 10.91743
Germany 10.89598
Table 11: Least Inquisitve
region opnScores
Macao SAR China 3.506623
Bhutan 3.642857
Cambodia 3.971429
Malaysia 4.221204
Nepal 4.783489
Gambia 5.300000

The Americas, Africa, and Europe seem to be a lot more open to new experiences, as they’re significantly brighter than the other continents. I can only think that risk-taking is just encouraged inside these societies or people are given more personal freedom and allowed more individuality. However, in other locations, they may prioritize stability and traditional methods – not necessarily a bad thing.

Are All Questions Created Equal?

We can also analyze the questions themselves. This quiz actually recorded how many milliseconds each respondent spent on each question.17

Thus, I present to you: Amount of time spent on each question!

For your reference:

I really like using a box plot18 to symbolize these results. For those who are unaware, the white box symbolizes the interquartile range (the range of values encompassing 25% to 75% of the data). The lower end of the box represents the bottom 25%, while the top end of the box represents the top 25% of users. The line in the center is the median value taken – in this case, the median time taken for each question. Box plots are great for visualizing the spread of data (which seems to range quite a bit) and managing outliars.

A note of the data: I’ve removed values above 20 seconds19 in the graph. I’ve also taken a sample of 50000 response times for each question.20

However, I want to be more precise when analyzing loading times, so I’ve specifically pulled the questions that take the longest and shortest times to complete.

Table 12: Questions That Took the Longest
Category Question Average Time in Seconds
AGR1_E I feel little concern for others. 17.050466
EXT2_E I don’t talk a lot. 14.906337
EST3_E I worry about things. 12.963980
CSN1_E I am always prepared. 10.388339
CSN4_E I make a mess of things. 9.143631
Table 12: Questions That Took the Shortest
Category Question Average Time in Seconds
EXT10_E I am quiet around strangers. 4.641537
EST9_E I get irritated easily. 4.620892
CSN9_E I follow a schedule. 4.529849
OPN8_E I use difficult words. 4.289319
OPN10_E I am full of ideas. 3.421338

I’ve taken the liberty of removing EXT_1 (the first question on the quiz), which had an average score of 87 seconds. I’d assume it’s skewed because people might’ve wanted to scroll around the webpage and get accustomed to it or maybe started then quiz then forgot about it.21

“I feel little concern for others” seems to be the question that takes the longest. It’s likely because it’s incredibly situational - ‘others’ is very vague. Does it mean if you care about everybody, including strangers or people you hate? Or should you just limit it to your friends, people you may feel an obligation to care about, or want to care about? This ‘situational’ factor also plagues the other high-ranking questions, where there are SOMETIMES you agree with the actions, and other times you don’t. There’s a lot of nuance that people need to consider, hence, requiring a longer response time.

On the other hand, the questions that took the least amount of time are concise and straightforward, contain simple words, and are obvious to judge yourself on. You either talk pretentiously to seem smart, or you don’t.22 You can also see that these questions are the ones that are asked later (i.e. A majority of the questions have an XXX9_E tag, meaning that it’s part of the 9th out of 10 rounds of questions, so they’d be the 40th - 45th question they’ve completed). Anticipation for results may have caused them to rush through the later questions. The reverse is also seen in the questions that look the longest (i.e. they show up earlier in the quiz).

Another interesting thing to note is the distribution of answers for each question. You may expected that each question has a distribution similar to standard distribution. However, that’s actually not the case! There are various patterns you can spot:

There are four major types of distribution seen in the data:

1. Logarithmic (First Row, i.e. I shirk my duties and I have a small heart)

The columns progressively increase or decrease. I think these graphs are the most “trustworthy”, as you’d only put 1 or 5 (the extremes) if you were incredibly confident in your answer.23 Not only that, but these questions aren’t really ‘shameful’ to admit, like “I get stressed out easily” is seen to be a relatively normal thing to say, which allows people to be honest and pick extremes.

Questions also tend to be less ‘situational’ and more ‘specific’, where you can very clearly visualize what you’d be doing in that situation, rather than responding “Oh… sometimes I am, sometimes I’m not!” (E.g. “I am quiet” likely wouldn’t follow this trend, but “I am quiet around strangers” does because the question narrow down the situation. You also don’t really change your behavior around different strangers - it’s pretty consistent.)

Some other examples include:

If you’re wondering which side they skew on…. just trust your gut on it :)

2. Skewed (Second Row, i.e. I am relaxed most of the time, I feel litte concern for others)

I feel like these are best associated with questions where self-bias is most prevalent, as you want to make yourself as something you’re not (to make yourself feel better). These questions are also very situational; Sometimes I do this, sometimes I do. That’s why people tend to lean towards the middle. This is the most common distribution type.

Examples include:

3. “Normal” Distribution (Third Row, Left Plot)

I’m lying to you - This graph isn’t normal distribution. I’m just calling it normal distribution because the answer 3 (the middle) is the most common answer – similar to a normal distribution graph. Quite frankly, I think this pattern means that people are either confused or they don’t really have a large opinion on it, so they’re almost forced to choose 3. These questions also seem to be the most ‘observable/objective’ of the bunch.

Personally, I try to make a habit on avoiding clicking 3 no matter what (I’m not sure if others are the same), but it’s still interesting to see. I feel like these things are not things to be ‘proud’ of or dislike about yourself, nor would you mention it unless prompted, which is likely why it’s typically associated with the extrovertism questions.

Examples Include:

4. Relatively Random (Third Row, Right)

These are the ones that don’t follow any of the trends above. Well, what happens is that 1 and 5 are similar in height and have about 150k responses, while 3, 4, 5 are also similar in height, with about 250k responses. Or, to generalize this pattern, it means that columns have relatively uniform distribution. It’s the kinds of questions that go “Yea… I’m definitely not SUPER ___, but it happens from time to time… I’m not sure how I would compare with other people though… so 2, 3, or 4 sound about right.”

Examples Include:


  1. Recall how introverted questions are negatively keyed and subtracted from your total.↩︎

  2. The average values will actually shift every time I update the website (on RStudio!), as the function ‘slice_sample’ will grab the average of 100 randomly selected responses, resulting in a unique average every time!↩︎

  3. https://www.wbur.org/hereandnow/2022/07/13/stress-survey-gallup↩︎

  4. https://www.voanews.com/a/why-people-worldwide-are-unhappier-more-stressed-than-ever-/6658784.html↩︎

  5. If someone told you, “Yea, you’re just a bit egocentric… no offense”, would you be happy about it?↩︎

  6. This was the only site that had theoretical values, but I don’t know where they got their percentages from. It should also be noted that this site does not use ‘X’ as a possible result. E.g. XLUEI (or any combination with an ‘X’) is not considered. So, this means that there is no data for specific combinations. Hence, the theoretical values should be trusted with a grain of salt. It should also be noted that SimilarMinds separates the theoretical values by female and male. Since this data set does not have this distinction, I used the average of the male and female theoretical values to get my average number.↩︎

  7. If you have more ideas, I’d love to hear them! Reach out :)↩︎

  8. There’s actually 3794 people who got XXXXX in this data set.↩︎

  9. ‘Country codes’ are kinda misleading. There are ~195 countries, and ISO has 249 different codes. This is because the ISO contains subdivisions of countries, e.g. Caymen Islands (UK) and Christmas Island (Australia).↩︎

  10. https://www.searchenginejournal.com/ranking-factors/physical-proximity-to-searcher↩︎

  11. I couldn’t find any specific location data on the website.↩︎

  12. As a Canadian, it’s actually the third search result!↩︎

  13. E.g. the 996 schedule of working from 9am to 9pm, 6 days a week.↩︎

  14. Youth in China are adopting the philosophy of ‘lying down’ and giving up, due to the bad job market after graduation. They often feel let down by their society, as they’ve always been taught that studying hard and getting into a good university will lead to a good life. Yet, when they graduate, they struggle to find jobs and keep themselves afloat.↩︎

  15. E.g. To get into university, students take the Gaokao, a two-day standardized test that determines their entire future.↩︎

  16. Report discussing link: https://www.econlib.org/archives/2009/03/econlog_book_cl_10.html Report of poverty in Sub-Saharan Africa: https://issafrica.org/iss-today/africa-is-losing-the-battle-against-extreme-poverty↩︎

  17. The timer starts when you answer to a previous question, and runs until you answer the current question. The first question’s timer starts when you load into the page. An ‘answer’ means clicking on of the options between 1 and 5.↩︎

  18. I actually forgot this type of graph existed, until I was randomly scrolling through the ggplot2 library of different graph types!↩︎

  19. As they’re significant outliars, and make the data difficult to visualize.↩︎

  20. 50000 is enough to show the general trend, and utilizing more only makes graphs take longer to load.↩︎

  21. Recall that the timer for question 1 starts as soon as the page loads (since there are no previous questions to start from).↩︎

  22. I’m referring to the difficult words question.↩︎

  23. I.e. If you ask people to choose a number between 1 and 5, they tend to pick 2, 3, or 4, rather than 1 or 5.↩︎

  24. This one is almost inverted, where 2 and 4 are the most common responses, followed by 1 and 5, then 3.↩︎